Enhanced Convolutional Neural Network for Efficient Content-Based Image Retrieval

Abstract

The use of picture objects in various real-world applications has increased dramatically with the rise of cloud-based ecosystems for managing, analyzing, and storing multimedia material. CBIR is a method for obtaining photos from the cloud and other storage infrastructures. It involves using an image input to look for images that match the database. Because of its methodology, this phenomenon is deemed preferable to text-based search. However, conventional CBIR techniques rely on similarity and feature comparison metrics. As AI grows, learning-based approaches are also shown to be beneficial for matching semantic material. Therefore, we presented a deep learning architecture to achieve an effective learning-based CBIR system in this research. To improve the matching experience in image retrieval, we suggested a modified CNN model for feature extraction from images. We proposed the Intelligent Content-Based Image Retrieval (ICBIR) algorithm. For our tests, we used the ImageNet micro dataset. The suggested modified CNN model-based CBIR system performs better than current techniques in picture retrieval that as closely resembles user intent as feasible, according to experimental data.